Messages in this thread | | | Date | Thu, 26 Feb 1998 10:02:11 -0700 (MST) | From | teunis <> | Subject | Re: 2.1.79, 2.1.87, 2.1.88 locks on EIDE (and a note on cdroms) |
| |
On Mon, 23 Feb 1998, Adam D. Bradley wrote:
> > > I have an FIC PA2007 mb with Cyrix 6x86 200+MX (75*2).... > > > > I'm getting random lockups under high disk activity with 2.1.79, 2.1.87, > > and 2.1.88.. > > > > The SCSI's don't trigger it - but EIDE does, whether I enable UDMA or not. > > I've been having problems with EIDE since 2.1.85 on my Gigabyte SMP > system (twin P166-MMX's). 2.1.84 seems rock-solid zero-problems, but > 2.1.85 and on have all been unreliable using either GCC 2.7.2.3 or > 2.8.0.
My problems appeared after 2.1.65 (though I'm [obviously] UP not SMP)
Update : I've found a fix (for my system):
Separate the UDMA drive from the Mode3 CDrom. Things seem stable now - I haven't been able to freeze it yet, even with several parallel fsck's on different filesystems.
On different channels the hard drive went into UDMA mode...
I suspect a linux-kernel bug, as 2.1.65 works perfectly (as does mickysloth's Win98). Prolly in the IDE section and prolly something to do with Mode 3 access.
Incidentally, the IDE cdrom interface _HAS_ to handle timeouts better. It takes a bit for this #^$& cdrom to spin up and during that time my logs fill up with timeout messages (as do my current consoles. This logging info is seriously irritating - guess I'll have to hunt it down). This has gotten _WORSE_ since 2.1.65 (where these messages first appeared on my system anyways). (the inability to mount a cdrom while the cdrom drive is spinning up is understandable though - but at least that command doesn't spit out multiline debug messages)
FWIW : Avoid Creative 24X cdroms. They work fine but they like to go to sleep when not used (and then cause timeouts while spinning back up).
G'day, eh? :) - Teunis
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |